The 3rd Generation Partnership Project (3GPP) is responsible for the standardization of the Universal Mobile Telecommunication System (UMTS) and Long Term Evolution (LTE). The 3GPP work on LTE is also referred to as Evolved Universal Terrestrial Access Network (E-UTRAN). LTE is a technology for realizing high-speed packet-based communication that can reach high data rates both in the downlink and in the uplink, and is thought of as a next generation mobile communication system relative to UMTS. In order to support high data rates, LTE allows for a system bandwidth of up to 20 MHz. LTE is also able to operate in different frequency bands and can operate in at least Frequency Division Duplex (FDD) and Time Division Duplex (TDD) modes. The modulation technique or the transmission method used in LTE is known as Orthogonal Frequency Division Multiplexing (OFDM). The first release of LTE is expected to provide peak rates of 300 Mbps, a radio-network delay of e.g. 5 ms or less, a significant increase in spectrum efficiency and a network architecture designed to simplify network operation, reduce cost, etc.
For the next generation mobile communications systems, e.g. International Mobile Telecommunications (IMT) advanced and/or LTE Advanced, which is an evolution of LTE, support for bandwidths of up to 100 MHz is being discussed. In both LTE and LTE Advanced, radio base stations are known as eNBs or eNodeBs, where “e” stands for evolved. Furthermore, multiple antennas with precoding and/or beamforming technology may be used in order to provide high data rates to user equipments (UEs). Thus, LTE and LTE Advanced both constitute examples of Multiple-Input, Multiple-Output (MIMO) radio systems. Another example of a MIMO- and OFDM based system is Worldwide Interoperability for Microwave Access (WiMAX). Since LTE Advanced is an evolution of LTE, backward compatibility is important so that LTE Advanced can be deployed in spectrum already occupied by LTE.
In LTE Advanced, also known as 3GPP Release 10, up to 8 layer transmission should be supported in order to fulfil LTE Advanced downlink spectral efficiency, 30 bps/Hz. This may be achieved by utilizing some kind of advanced antenna configuration, e.g. 8×8 high-order MIMO, where 8 transmit antennas and 8 receive antennas are used. Throughout this document, the term “antenna port” will be used rather than antenna, to emphasize that what is referred to does not necessarily correspond to a single physical antenna.
To provide context for the subsequent disclosure, a brief review of the LTE downlink physical resource structure will now be provided. In OFDM systems such as LTE, the available physical resources are divided into a time and frequency grid. The time dimension is divided into subframes, each comprising a number of OFDM symbols. In LTE and LTE Advanced, a subframe is 1 ms in length, divided into two time slots of 0.5 ms each. A guard interval, called a cyclic prefix (CP), is prepended to each OFDM symbol in order to reduce inter-symbol interference. For normal cyclic prefix (CP) length, the number of OFDM symbols per subframe is 14, which implies that time is quantized into 14 symbols during a subframe. For extended cyclic prefix length, there are 12 OFDM symbols per subframe. Frequency corresponds to subcarriers in the OFDM symbols, and the number of subcarriers varies depending on the system bandwidth used. Each box within the time-frequency grid represents a single subcarrier for one symbol period, and is referred to as a resource element. The smallest schedulable unit of resource elements is called a physical resource block (PRB), or simply a resource block (RB). In LTE and LTE Advanced, a resource block spans 12 subcarriers and 0.5 ms, i.e. 7 or 6 OFDM symbols depending on cyclic prefix length. The resource blocks are, however, allocated in pairs in the time domain. Thus, an LTE subframe of 1 ms is two resource blocks wide.
There is also a special type of LTE subframe, composed of three fields: Downlink Pilot Timeslot (DwPTS), Guard Period (GP), and Uplink Pilot Timeslot (UpPTS). This special subframe is used for downlink-to-uplink switching in TDD mode. The duration of the GP field is varied depending on how long it takes the UE to switch from receiving to sending, and also on the signal propagation time from the base station to the UE. The DwPTS field carries synchronization and user data, as well as the downlink control channel for transmitting scheduling and control information. Since the total subframe duration is fixed at 1 ms, the duration of the DwPTS and UpPTS fields are adjusted based on the duration of the GP field.
A reference signal is a known signal which is inserted at predetermined positions in the OFDM time-frequency grid. The presence of this known signal allows the UE to estimate the downlink channel so that it may carry out coherent channel demodulation. It has been agreed for LTE that up to 8 UE-specific reference signals (RS) will be introduced for the purpose of channel demodulation. The UE-specific reference signals are also called demodulation RS or DM-RS. Thus, each antenna port transmits one DM-RS, which is specific to that antenna port as well as to the UE that the transmission is directed to.
Reference signals are generally transmitted according to a predefined pattern in time and frequency, so that the UE knows where to find the signals. A prior art DM-RS pattern with normal cyclic prefix (CP), supporting up to rank 8, is shown in FIG. 1. The expression “rank”, or transmission rank, refers to the number of independent data streams, or spatial layers, which may be reliably transmitted over a wireless channel. In the present context, the rank may be interpreted as the maximum number of transmit antenna ports that are supported.
FIG. 1 shows a time-frequency grid for a normal subframe, i.e. not a special subframe. Each row in the grid represents a subcarrier, and each column represents an OFDM symbol. The first three OFDM symbols are rendered in light gray color, to indicate that these symbols may be reserved for control signalling. The grid covers two LTE time slots, as explained above. The DM-RS pattern of FIG. 1 supports a total of 8 DM-RS antenna ports. The pattern exhibits a DM-RS overhead of 12 RE per layer; that is to say, each antenna port will use 12 REs per subframe for transmitting reference signals. For instance, one antenna port will transmit reference signals in the REs represented by the 12 squares filled with slanted lines in FIG. 1. The 8 DM-RS antenna ports are separated by a combination of CDM and FDM, as will be further explained below. It should be understood that other kinds of reference signals may also be transmitted; however, these have been omitted from FIG. 1 for reasons of simplicity.
Up to two code division multiplexing (CDM) groups are reserved for DM-RS, where each CDM group consists of 12 resource elements (RE) per physical resource block (PRB) pair. In the context of this disclosure, a CDM group is a group of resource elements which are used for multiplexing reference signals from a number of antenna ports using code division multiplexing. Thus, the 12 squares with slanting lines in FIG. 1 form one CDM group, and the 12 squares with horizontal lines form another CDM group. Each CDM group supports a maximum of four layers, i.e. a maximum of four antenna ports. The two CDM groups are multiplexed by FDM; in other words, the REs belonging to the first and second CDM groups are transmitted on different frequencies, i.e. subcarriers.
There is one CDM cluster in each time slot, as indicated by the thick black outlines 110, 120 in FIG. 1.
Furthermore, each CDM group comprises three CDM subgroups, i.e. groups of resource elements that share the same subcarrier. For example, the four squares with slanted lines in the top row of the time-frequency grid in FIG. 1 form one CDM subgroup, as indicated by the thick gray outline 130. Two further subgroups are indicated by thick gray outlines 140 and 150. Each CDM subgroup comprises 4 REs in the time domain, and in each CDM subgroup, up to four DM-RS antenna ports may be multiplexed.
The multiplexing of reference signals within a CDM subgroup is achieved by applying orthogonal cover codes (OCC) across the time domain. An OCC is a set of codes which all have zero cross-correlation. Thus, two signals encoded with two different codes from the set will not interfere with one another. An example of an OCC is a Walsh code. Walsh codes are defined using a Walsh matrix of length N, i.e. having N columns. Each row in the Walsh matrix is one length-N Walsh code. For example, the Walsh matrix of length-4 is:
  W  =      (                            1                          1                          1                          1                                      1                                      -            1                                    1                                      -            1                                                1                          1                                      -            1                                                -            1                                                1                                      -            1                                                -            1                                    1                      )  
Each row in this matrix forms one code of length 4, i.e. the codes are [1, 1, 1, 1], [1, −1, 1, −1], [1, 1, −1, −1] and [1, −1, −1, 1]. These four codes are all orthogonal with respect to each other. The individual “1”:s and “−1”:s of each code will be referred to as “code elements” in the following.
Although Walsh codes will be used throughout this disclosure to exemplify the invention, it should be understood that any OCC may be used. When this disclosure refers to “applying an orthogonal cover code” or “transmitting a signal using an orthogonal cover code” this should be understood as referring to one code out of a set of mutually orthogonal codes, e.g. one row from the Walsh matrix.
Each antenna port transmits one reference signal within the CDM subgroup, by applying an orthogonal cover code to the signal. If four antenna ports are multiplexed within a CDM subgroup, a length-4 OCC will be used, and each of the four antenna ports will use a different code from the set. This allows the reference signals to be separated and decoded on the receiver side.
The concept of OCC mapping has been introduced for dual layer beamforming, with the aim to reach full peak power randomization, which is expected to improve eNodeB side power utilization. OCC mapping means that the code elements in each OCC are mapped to reference elements in a specific pattern, or a specific order. One example of an OCC mapping design, which uses length-2 Walsh codes, is shown in FIG. 2. In the lower right corner of FIG. 2, the length-2 Walsh matrix is shown. Since a length-2 code is used, two antenna ports are multiplexed in each CDM subgroup in this example. Each antenna port will transmit two reference signals; one in the first time slot, and one in the second time slot. Layer 1, i.e. the first antenna port, uses the code from the first row in the Walsh matrix, i.e. [+1, +1]. Layer 2, i.e. the second antenna port, uses the code from the second row, [+1, −1]. Index a corresponds to the first code element, and index b corresponds to the second code element of each code. Thus, in the second code [+1, −1], index a corresponds to +1 and b corresponds to −1. Each antenna port will encode its reference signal by applying the code elements in the order indicated by the pattern of a:s and b:s in the time-frequency grid of FIG. 2.
An example may help illustrate the encoding process. Focusing on the first CDM subgroup 210, the first antenna port will transmit two reference signals, denoted X1 and X2, in this CDM subgroup. The second antenna port will also transmit two reference signals, denoted Y1 and Y2, in the same CDM subgroup 210. The first antenna port will encode its first reference signal, X1, in OFDM symbols 6 and 7 by applying the code elements [a, b], corresponding to [+1, +1], since the first antenna port uses the first Walsh code. Thus, the first antenna port will transmit [X1, X1]. The second antenna port will also encode its first reference signal, denoted Y1, in OFDM symbols 6 and 7. It will apply the code elements [a, b] from the second Walsh code, i.e. [+1, −1]. Therefore, the second antenna port will transmit [Y1, −Y1]. These signals will be superimposed, so that the resulting signal transmitted in OFDM symbols 6 and 7 is [X1+Y1, X1−Y1].
However, in the second CDM subgroup 220, i.e. the sixth row of the time-frequency grid, the two antenna port will encode their reference signals by applying the code elements in reverse order. Focusing again on OFDM symbols 6 and 7, the first antenna port will use the code [+1, +1], i.e. [X1, X1]—effectively the same code again, as reversing the code elements makes no difference in this case—but the second antenna port will use the code [−1, +1], i.e. [−Y1, Y1]. Thus, the resulting signal transmitted in OFDM symbols 6 and 7 in the second CDM subgroup 220 will be [X1−Y1, X1+Y1].
For completeness, it is pointed out that each antenna port will also transmit a second reference signal, denoted X2 and Y2, respectively, in OFDM symbols 13 and 14. The code pattern is the same as in the previous example and the resulting signal transmitted in OFDM symbols 13 and 14 may be derived in the same way.
It is pointed out that in this example, only CDM group 1 is allocated. Also, the mapping pattern is different in even PRBs and odd PRBs. Full peak power randomization can be reached between two adjacent PRBs. To understand why, consider the special case where reference signals X1 and Y1 are the same, i.e. X1=X2. Using the same example as above, the signal transmitted in symbols 6 and 7 of the first CDM subgroup 210 will be [X1+X1, X1−X1], i.e. [2X1, 0]. In the second CDM subgroup 220, the resulting signal will be [X1−X1, X1+X1], i.e. [0, 2X1]. Thus, in OFDM symbol 6, the signal 2X1 will be transmitted in the first CDM subgroup 210, and 0 will be transmitted in the second CDM subgroup 220. In OFDM symbol 7, the situation is the reverse, i.e. 0 in the first CDM subgroup 210, and 2X1 in the second CDM subgroup 220. This means that the total transmit power level will be about the same in OFDM symbol 6 as in symbol 7. In other words, the transmit power level is balanced between OFDM symbols, which implies that high peaks in transmit power levels between symbols may be avoided.
As mentioned above, the use of orthogonal cover codes enables the receiver to decode the reference signals in order to estimate the channel. Thus, at the UE side, per port channel estimation is performed by using the proper OCC. In other words, each reference signal is decoded, or despread, using the corresponding OCC that was used to encode the signal. A different length OCC is applied for channel estimation depending on how many layers are multiplexed in one CDM group. Two example cases with two and four layers, respectively, will now be described with reference to FIGS. 3(a) and 3(b).                When up to two layers are multiplexed in one CDM group, a length-2 OCC can be used for each CDM cluster 340, 350 in both slots, as shown in FIG. 3(a). This means that the Doppler impact introduced by mobility can be well captured by weighting two CDM clusters.        When more than two layers are multiplexed in one CDM group, a length-4 OCC has to be used across both clusters in one subframe, as illustrated in FIG. 3(b). Length-4 OCC is typically used for high rank cases, i.e. four or more antenna ports.        
At the UE side, one common strategy for performing DM-RS based channel estimation is to apply a 2×1D filter method per PRB, i.e. first a frequency domain filter and then a time domain filter. The basic principle is shown in FIG. 4. Frequency domain filtering and time domain filtering are performed based on respective inputs of delay spread, Doppler, and SNR. Due to uncertain resource allocation and bandwidth, the frequency domain filter has been found to require a much longer processing time than the time domain filter. To some extent, the time required by the frequency domain filter becomes a bottleneck which prevents speeding up the processing on channel estimation and further detection, and this may impact the overall detection latency.
When performing channel estimation with a length-2 OCC, as shown in FIG. 3(a), we notice that slot-by-slot channel estimation can be exploited. That is to say, channel estimation in the 1st slot can be performed first before the reception of the whole subframe. The reason for this is that a reference signal is transmitted in two consecutive REs, which are comprised in the same time slot. In other words, all the information required to decode the reference signal is available within a single time slot. This allows the processing time taken by the frequency domain filter in the first slot to be reduced, since the information received in the first slot can be processed during the time the second slot is received. This may result in a low latency channel estimator.
However, in 3GPP Release 10, a length-4 OCC is used to support multiplexing of up to four layers in each CDM group, as has been explained above. When performing channel estimation with length-4 OCC, as shown in FIG. 3(b), a length-4 OCC is used instead of length-2 OCC. However, length-4 OCC despreading cannot be performed until the whole subframe is received. This is because each reference signal is spread across four REs, which are distributed across two time slots (see FIG. 1). Thus, in the conventional scheme, channel estimation cannot be performed until both time slots are received. This means that processing of the first slot can not be performed in parallel with receiving the second slot, and additional time will be required, particularly by the frequency domain filter. Consequently, there is a risk of higher latency when performing channel estimation in the length-4 OCC case, since slot-by-slot channel estimation is not possible, as for the length-2 OCC case. In addition, in case of a length-4 OCC, the Doppler impact can not be well overcome since code despreading needs to be considered in both slots.
Furthermore, the OCC mapping pattern shown in FIG. 2 achieves full peak power randomization over two RBs, as described above, but only for normal cyclic prefix (CP) length. Thus, there is a need for a mechanism for enabling full peak power randomization also in the extended CP case, and/or for special subframes comprising the DwPTS field (Downlink Pilot Timeslot).